Method, system, device and readable storage medium for grading proximal maculopathy

Abstract

The present invention provides a grading method, system, device and readable storage medium for approximate macular lesions, the grading method comprising: obtaining a fundus image to be graded; preprocessing the fundus image to be graded to obtain a processed fundus image; the processed fundus image meets the preset grading processing requirements; inputting the processed fundus image into a pre-stored myopic macular lesion grading model, and grading the lesions of the processed fundus image by the pre-stored myopic macular lesion grading model to output a predicted result of the degree of pathological myopia. The grading method, system, device and readable storage medium for approximate macular lesions of the present invention realize the grading of myopic macular lesions, with high grading efficiency and high accuracy.

Description

Technical Field

[0001] The present invention belongs to the field of image processing technology, and relates to a grading method and system, and in particular to a grading method, system, device and readable storage medium for proximal maculopathy. Background Art

[0002] With the development of society and the continuous improvement of people's living standards, vision problems caused by myopia and high myopia have become increasingly prominent. Pathological myopia is a disease that is based on simple myopia, with obvious impairment of visual function, unsatisfactory correction of distance vision, varying degrees of fundus lesions, and may be accompanied by degenerative lesions such as posterior scleral staphyloma. About 3% of the world's population suffers from pathological myopia, which is one of the three major blinding diseases in the world. It is especially common in Asians and mostly occurs in young and middle-aged people. Pathological myopia generally has more characteristic fundus macular lesions. Previous studies by Ohno-Matsui et al. divided their macular lesions into 5 levels, as shown in Table 1. This is also the grading standard widely used in clinical practice. This grading result reflects the severity of pathological myopia to a certain extent and can guide clinical treatment.

[0003] At present, fundus camera imaging has become one of the most commonly used clinical ophthalmic examination methods because of its simplicity and convenience. Fundus images can be used to make a preliminary judgment on whether there is a lesion in the fundus. However, when doctors observe fundus images for diagnosis, there are shortcomings such as large workload, poor repeatability, and too strong subjectivity in judgment criteria. Therefore, some machine learning methods such as deep learning have been used to analyze and process fundus images to assist doctors in diagnosis.

[0004] However, the existing technology only performs binary classification on fundus images, that is, judging whether a certain lesion occurs, while ignoring the subdivision of the degree and level of the lesion. At the same time, for pathological myopia, a fundus disease that has gradually received attention in recent years, there is currently a lack of mature deep learning-based detection and screening technology.

[0005] Therefore, how to provide a grading method, system, device and readable storage medium for proximal macular degeneration to solve the defects of the existing technology, such as the lack of mature deep learning-based detection and screening methods, which leads to the neglect of the subdivision of the degree and level of the lesions, has become a technical problem that needs to be urgently solved by technical personnel in this field. Summary of the invention

[0006] In view of the shortcomings of the prior art mentioned above, the purpose of the present invention is to provide a grading method, system, device and readable storage medium for proximal macular degeneration, so as to solve the problem that the prior art lacks a mature deep learning-based detection and screening method, resulting in neglect of the subdivision of the degree and level of the lesion.

[0007] To achieve the above-mentioned purpose and other related purposes, the present invention provides a method for grading myopic macular lesions, comprising: obtaining a fundus image to be graded; preprocessing the fundus image to be graded to obtain a processed fundus image, wherein the processed fundus image meets preset grading processing requirements; inputting the processed fundus image into a pre-stored myopic macular lesion grading model, and grading the lesions of the processed fundus image using the pre-stored myopic macular lesion grading model to output a predicted result of the degree of pathological myopia.

[0008] In one embodiment of the present invention, the preprocessing of the fundus image to be graded includes: cropping the width and height of the fundus image to be graded to be consistent; normalizing the cropped fundus image so that the pixel values ​​in the fundus image are distributed between 0 and 1, thereby forming a processed fundus image.

[0009] In one embodiment of the present invention, the pre-stored myopic maculopathy grading model includes a residual network, a fully connected layer and a classification layer; wherein the residual network is composed of a convolutional layer and a convolutional layer group.

[0010] In one embodiment of the present invention, the step of grading the lesions of the processed fundus image using the pre-stored myopic maculopathy grading model includes: extracting abstract features related to approximate maculopathy in the processed fundus image through the residual network to form a feature map related to approximate maculopathy; and mapping the feature map related to approximate maculopathy to the lesion grading space through the classification layer.

[0011] In one embodiment of the present invention, the step of grading the lesions of the processed fundus image using the myopic maculopathy grading model further includes: converting the feature vectors in the feature map related to the myopic maculopathy through the fully connected layer.

[0012] In one embodiment of the present invention, the lesion grading space includes a grading space corresponding to leopard-like fundus, a grading space corresponding to diffuse choroidal atrophy, a grading space corresponding to patchy choroidal atrophy, and a grading space corresponding to macular atrophy.

[0013] In one embodiment of the present invention, after extracting the abstract features related to approximate maculopathy, the step of grading the lesions of the processed fundus image using the myopic maculopathy grading model also includes: performing a maximum pooling operation on the feature map related to approximate maculopathy.

[0014] On the other hand, the present invention provides a grading system for myopic maculopathy, including: an image acquisition module for acquiring a fundus image to be graded; a preprocessing module for preprocessing the fundus image to be graded to acquire a processed fundus image, wherein the processed fundus image meets preset grading processing requirements; a grading module for inputting the processed fundus image into a pre-stored myopic maculopathy grading model, and grading the lesions of the processed fundus image using the pre-stored myopic maculopathy grading model to output a predicted result of the degree of pathological myopia.

[0015] Yet another aspect of the present invention provides a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for grading proximal maculopathy.

[0016] The last aspect of the present invention provides a grading device for proximal macular lesions, comprising: a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the grading device for proximal macular lesions executes the grading method for proximal macular lesions.

[0017] As described above, the method, system, device and readable storage medium for grading proximal maculopathy of the present invention have the following beneficial effects:

[0018] The grading method, system, device and readable storage medium for myopic maculopathy described in the present invention realize the grading of myopic maculopathy with high grading efficiency and high accuracy, filling the gap in the market. According to this grading, it is possible to provide strong technical auxiliary support for the diagnosis and treatment of pathological myopia, a disease that is increasingly concerned in clinical practice, help doctors make decisions quickly and effectively, and also provide intelligent screening and auxiliary diagnosis services in primary medical and health institutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 FIG. 1 is a flow chart of a method for grading proximal maculopathy according to an embodiment of the present invention.

[0020] Figure 2 Shown is a schematic structural diagram of the pre-stored myopic maculopathy grading model of the present invention.

[0021] Figure 3 It is a schematic diagram of the process of S13 of the present invention.

[0022] Figure 4 Shown is a schematic diagram of the BottleNeck structure in each convolutional layer group in Resnet-101 of the present invention.

[0023] Figure 5It is a schematic diagram showing the principle structure of a method for grading proximal maculopathy of the present invention in one embodiment.

[0024] Component number description

[0025] 5 Grading system for proximal maculopathy

[0026] 51 Image acquisition module

[0027] 52 Preprocessing module

[0028] 53 Grading Modules

[0029] Steps S11 to S13

[0030] Steps S131 to S134 DETAILED DESCRIPTION

[0031] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0032] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and thus the drawings only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0033] Embodiment 1

[0034] This embodiment provides a method for grading proximal maculopathy, including:

[0035] Acquiring a fundus image to be graded;

[0036] Preprocessing the fundus image to be graded to obtain a processed fundus image, wherein the processed fundus image meets a preset grading processing requirement;

[0037] The processed fundus image is input into a pre-stored myopic macular lesion grading model, and the processed fundus image is graded by the pre-stored myopic macular lesion grading model to output a prediction result of the degree of pathological myopia.

[0038] The following will describe in detail the grading method of proximal maculopathy provided by the embodiment in conjunction with the diagram. Figure 1 , which is a flow chart of a method for grading proximal maculopathy in one embodiment. Figure 1 As shown, the grading method of proximal maculopathy specifically includes the following steps:

[0039] S11, obtaining a fundus image to be graded.

[0040] S12, pre-processing the fundus image to be graded to obtain a processed fundus image. The processed fundus image meets the preset graded processing requirements. In this embodiment, the preset graded processing requirements include requiring the image width and height to be consistent, and requiring the pixel values ​​in the image to be distributed between 0 and 1.

[0041] Specifically, the S12 includes:

[0042] The width and height of the fundus image to be graded are cut to be consistent. For example, the fundus image to be graded is cut to 224×224.

[0043] The cropped fundus image is normalized so that the pixel values ​​in the fundus image are distributed between 0 and 1, thereby forming a processed fundus image.

[0044] S13, inputting the processed fundus image into a pre-stored myopic macular lesion grading model, and grading the lesions of the processed fundus image by the myopic macular lesion grading model to output a prediction result of the degree of pathological myopia.

[0045] See also Figure 2 , which is a schematic diagram of the structure of the pre-stored myopic maculopathy grading model. Figure 2 As shown, the pre-stored myopic maculopathy grading model includes a residual network (in this embodiment, specifically a Resnet-101 network), a fully connected layer and a classification layer (in this embodiment, specifically a Softmax classification layer); wherein the residual network is composed of a convolutional layer and a convolutional layer group.

[0046] In this embodiment, the training process of the pre-stored myopic maculopathy grading model is as follows:

[0047] First, original fundus images are collected as training datasets.

[0048] The training data set is collected from a certain hospital, and contains a total of 16,651 fundus images with lesion severity levels annotated.

[0049] Among them, 10,216 images are of normal + other lesions, 4,220 images are of leopard-marked fundus, 1,154 images are of diffuse choroidal atrophy, 678 images are of patchy choroidal atrophy, and 383 images are of macular atrophy. The annotations of the fundus images are all from professional ophthalmologists.

[0050] Next, divide the set into training and test sets.

[0051] The training set and test set are divided into a ratio of 4:1. Since the proportion of samples in each category in the data set is different, the samples are randomly divided in each category at a ratio of 4:1 to ensure that the proportion of samples in each category in the final training set and test set is basically consistent.

[0052] Finally, train the network model.

[0053] The network is trained using the training set, and the network parameters are all randomly initialized.

[0054] Since the number of samples of each category in the training set is unevenly distributed, in this embodiment, the loss function selects the weighted cross entropy loss function to reduce the impact of the category imbalance problem on the model training effect. The specific weighted cross entropy loss function is as follows:

[0055]

[0056] Where N represents the number of samples. M = 5 represents the number of categories. c Represents the weight of this category in the loss function calculation. ic is an indicator scalar (0 or 1), which is 1 if the category c is the same as the true category of the sample, otherwise it is 0. ic It represents the probability that sample i belongs to category C.

[0057] The initial learning rate is set to 0.001, and the cosine annealing learning rate decay strategy is used to change the learning rate. The stochastic gradient descent algorithm is used, and all data in the training set pass through the network once to complete one round. After the network converges, the network parameters are fixed and the training is completed.

[0058] In this example, see Figure 3 , which is a schematic diagram of the process of S13. Figure 3 As shown, the S13 includes:

[0059] S131, extracting abstract features related to approximate macular degeneration in the processed fundus image through the residual network to form a feature map related to approximate macular degeneration.

[0060] The Resnet-101 network includes the first convolutional layer, the second convolutional layer group, the third convolutional layer group, the fourth convolutional layer group and the fifth convolutional layer group. The first convolutional layer is a two-dimensional convolution with a convolution kernel size of 7×7, followed by BatchNorm and ReLU. The second to fifth convolutional layer groups are respectively composed of a certain number of BottleNecks, and there is a MaxPooling layer before the second convolutional layer group.

[0061] S132, after extracting the abstract features related to the approximate macular lesions in the processed fundus image, a maximum pooling operation is performed on the feature map related to the approximate macular lesions through a MaxPooling layer. The maximum pooling operation can reduce the image resolution, that is, reduce the image length and width to one-half.

[0062] See also Figure 4 , which shows the BottleNeck structure diagram in each convolutional layer group in Resnet-101. Figure 4 As shown, the BottleNeck is composed of Conv1×1, BatchNorm, ReLU, Conv3×3, BatchNorm, ReLU, Conv1×1, BatchNorm, and ReLU connected in sequence. The Conv1×1 operation is used to extract the abstract features of the image. The BatchNorm layer works in the training process, which normalizes the training data to reduce data jitter and speed up training, but this layer does not work in the prediction process.

[0063] S133, transforming the feature vector in the feature map related to proximal maculopathy through the fully connected layer to obtain a transformed vector.

[0064] For example, a feature map of size 2048×1×1 is treated as a 2048×1 feature vector and input into the fully connected layer, and the output is a 5×1 vector.

[0065] S134, mapping the feature map related to the approximate macular lesion (specifically, the converted vector) to the lesion classification space through the Softmax classification layer. In this embodiment, the lesion classification space includes a grade space corresponding to leopard-like fundus, a grade space corresponding to diffuse choroidal atrophy, a grade space corresponding to patchy choroidal atrophy, and a grade space corresponding to macular atrophy.

[0066] This embodiment also provides a readable storage medium (also called a computer-readable storage medium) on which a computer program is stored. When the program is executed by a processor, the above-mentioned grading method for proximal maculopathy is implemented.

[0067] A person skilled in the art can understand that computer-readable storage media: all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to the computer program. The aforementioned computer program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk and other media that can store program codes.

[0068] The grading method for myopic maculopathy described in this embodiment realizes the grading of myopic maculopathy with high grading efficiency and high accuracy, filling the gap in the market. According to this grading, it can provide strong technical auxiliary support for the diagnosis and treatment of pathological myopia, a disease that has received increasing attention in clinical practice, help doctors make decisions quickly and effectively, and also provide intelligent screening and auxiliary diagnosis services in primary medical and health institutions.

[0069] Embodiment 2

[0070] This embodiment provides a grading system for proximal maculopathy, including:

[0071] An image acquisition module, used for acquiring fundus images to be graded;

[0072] A preprocessing module, used for preprocessing the fundus image to be graded to obtain a processed fundus image, wherein the processed fundus image meets the preset grading processing requirements;

[0073] The grading module is used to input the processed fundus image into a pre-stored myopic macular lesion grading model, and perform lesion grading on the processed fundus image through the pre-stored myopic macular lesion grading model to output a predicted result of the degree of pathological myopia.

[0074] The grading system for proximal maculopathy provided by this embodiment will be described in detail below with reference to the diagram. Figure 5 , which is a schematic diagram showing the principle structure of a grading system for proximal maculopathy in one embodiment. Figure 5 As shown, the proximal maculopathy grading system 5 includes an image acquisition module 51 , a preprocessing module 52 and a grading module 53 .

[0075] The image acquisition module 51 is used to acquire the fundus image to be graded.

[0076] The preprocessing module 52 is used to preprocess the fundus image to be graded and obtain a processed fundus image that meets the preset graded processing requirements. In this embodiment, the preset graded processing requirements include requiring the image width and height to be consistent and requiring the pixel values ​​in the image to be distributed between 0 and 1.

[0077] Specifically, the preprocessing module 52 crops the width and height of the fundus image to be classified to be consistent, and normalizes the cropped fundus image so that the pixel values ​​in the fundus image are distributed between 0 and 1, thereby forming a processed fundus image.

[0078] The grading module 53 is used to input the processed fundus image into a pre-stored myopic maculopathy grading model, and perform lesion grading on the processed fundus image using the pre-stored myopic maculopathy grading model to output a prediction result of the degree of pathological myopia.

[0079] In this embodiment, in this embodiment, the pre-stored myopic maculopathy grading model includes a residual network (in this embodiment, specifically a Resnet-101 network), a fully connected layer and a classification layer (in this embodiment, specifically a Softmax classification layer); wherein the residual network is composed of a convolutional layer and a convolutional layer group.

[0080] The Resnet-101 network includes the first convolutional layer, the second convolutional layer group, the third convolutional layer group, the fourth convolutional layer group and the fifth convolutional layer group. The first convolutional layer is a two-dimensional convolution with a convolution kernel size of 7×7, followed by BatchNorm and ReLU. The second to fifth convolutional layer groups are respectively composed of a certain number of BottleNecks, and there is a MaxPooling layer before the second convolutional layer group.

[0081] The BottleNeck is composed of Conv1×1, BatchNorm, ReLU, Conv3×3, BatchNorm, ReLU, Conv1×1, BatchNorm, and ReLU connected in sequence. The Conv1×1 operation is used to extract the abstract features of the image. The BatchNorm layer works during the training process, which normalizes the training data to reduce data jitter and speed up training, but this layer does not work during the prediction process.

[0082] Specifically, the grading module 33 extracts abstract features related to approximate macular lesions in the processed fundus image through the residual network to form a feature map related to approximate macular lesions; and performs a maximum pooling operation on the feature map related to approximate macular lesions through the MaxPooling layer. The feature vector in the feature map related to approximate macular lesions is converted through the fully connected layer to obtain the converted vector. The feature map related to approximate macular lesions (specifically, the converted vector) is mapped to the lesion grading space through the Softmax classification layer. In this embodiment, the lesion grading space includes a grade space corresponding to a leopard-like fundus, a grade space corresponding to diffuse choroidal atrophy, a grade space corresponding to patchy choroidal atrophy, and a grade space corresponding to macular atrophy.

[0083] It should be noted that it should be understood that the division of the various modules of the above system is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. And these modules can all be implemented in the form of software called by processing elements, or they can all be implemented in the form of hardware, or some modules can be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example: the x module can be a separately established processing element, or it can be integrated in a certain chip of the above system. In addition, the x module can also be stored in the memory of the above system in the form of program code, and called and executed by a certain processing element of the above system. The implementation of other modules is similar. These modules can be integrated together in whole or in part, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in the processor element or instructions in the form of software. The above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), one or more digital singnal processors (DSP), one or more field programmable gate arrays (FPGA), etc. When a module is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. These modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0084] Embodiment 3

[0085] The present embodiment provides a device for grading proximal macular lesions, comprising: a processor, a memory, a transceiver, a communication interface and / or a system bus; the memory and the communication interface are connected to the processor and the transceiver via the system bus and communicate with each other, the memory is used to store computer programs, the communication interface is used to communicate with other devices, and the processor and the transceiver are used to run computer programs, so that the device for grading proximal macular lesions performs each step of the method for grading proximal macular lesions as described in Example 1.

[0086] The system bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus. The communication interface is used to realize the communication between the database access device and other devices (such as clients, read-write libraries, and read-only libraries). The memory may include random access memory (RAM), and may also include non-volatile memory, such as at least one disk storage.

[0087] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0088] The protection scope of the method for grading proximal macular degeneration described in the present invention is not limited to the execution order of the steps listed in this embodiment. All solutions implemented by adding, reducing or replacing steps in the prior art based on the principles of the present invention are included in the protection scope of the present invention.

[0089] The present invention also provides a grading system for proximal macular lesions, which can implement the grading method for proximal macular lesions described in the present invention. However, the implementation device of the grading method for proximal macular lesions described in the present invention includes but is not limited to the structure of the large amount of data access system listed in this embodiment. All structural deformations and replacements of the prior art made according to the principles of the present invention are included in the protection scope of the present invention.

[0090] In summary, the grading method, system, device and readable storage medium of myopic maculopathy of the present invention realizes the grading of myopic maculopathy, has high grading efficiency and high precision, fills the gap in the market, and can provide powerful technical auxiliary support for the diagnosis and treatment of pathological myopia, a disease that has received increasing attention, in the clinic, according to this grading, help doctors make decisions quickly and effectively, and can also provide intelligent screening and auxiliary diagnosis services in primary medical and health institutions. The present invention effectively overcomes various shortcomings in the prior art and has high industrial utilization value.

[0091] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A method for grading proximal maculopathy, characterized in that: include: Acquiring a fundus image to be graded; Preprocessing the fundus image to be graded to obtain a processed fundus image, wherein the processed fundus image meets a preset grading processing requirement; Inputting the processed fundus image into a pre-stored myopic maculopathy grading model, the pre-stored myopic maculopathy grading model includes a residual network, a fully connected layer and a classification layer; wherein the residual network is composed of a convolutional layer and a convolutional layer group, the residual network includes a Resnet-101 network, the Resnet-101 network includes a first convolutional layer, a second convolutional layer group, a third convolutional layer group, a fourth convolutional layer group and a fifth convolutional layer group, and the first convolutional layer is followed by BatchNorm; The loss function used when training the pre-stored myopic maculopathy grading model is a weighted cross entropy loss function; The processed fundus image is graded for lesions using the pre-stored myopic macular lesion grading model, including: extracting abstract features related to approximate macular lesions in the processed fundus image through the residual network to form a feature map related to approximate macular lesions; mapping the feature map related to approximate macular lesions to a lesion grading space through the classification layer to output a prediction result of the degree of pathological myopia, wherein the lesion grading space includes a grade space corresponding to a leopard-like fundus, a grade space corresponding to diffuse choroidal atrophy, a grade space corresponding to patchy choroidal atrophy, and a grade space corresponding to macular atrophy.

2. The method for grading proximal maculopathy according to claim 1, characterized in that: The preprocessing of the fundus image to be graded comprises: Cutting the width and height of the fundus image to be graded to be consistent; The cropped fundus image is normalized so that the pixel values ​​in the fundus image are distributed between 0 and 1, thereby forming a processed fundus image.

3. The method for grading proximal maculopathy according to claim 1, characterized in that: The step of performing lesion grading on the processed fundus image by using the myopic maculopathy grading model further comprises: The feature vector in the feature map related to proximal maculopathy is transformed through the fully connected layer.

4. The method for grading proximal maculopathy according to claim 1, characterized in that: After extracting the abstract features related to myopic maculopathy, the step of grading the lesions on the processed fundus image using the myopic maculopathy grading model further includes: A maximum pooling operation is performed on the feature maps associated with proximal maculopathy.

5. A grading system for proximal maculopathy, characterized in that: include: An image acquisition module, used for acquiring fundus images to be graded; A preprocessing module, used for preprocessing the fundus image to be graded to obtain a processed fundus image, wherein the processed fundus image meets the preset grading processing requirements; A grading module is used to input the processed fundus image into a pre-stored myopic maculopathy grading model, wherein the pre-stored myopic maculopathy grading model includes a residual network, a fully connected layer and a classification layer; wherein the residual network is composed of a convolutional layer and a convolutional layer group, and the residual network includes: a Resnet-101 network, wherein the Resnet-101 network includes a first convolutional layer, a second convolutional layer group, a third convolutional layer group, a fourth convolutional layer group and a fifth convolutional layer group, and the first convolutional layer is followed by BatchNorm; the loss function used when training the pre-stored myopic maculopathy grading model is a weighted cross entropy loss Function; grading the lesions of the processed fundus image through the pre-stored myopic macular lesion grading model, extracting the abstract features related to the approximate macular lesions in the processed fundus image through the residual network, and forming a feature map related to the approximate macular lesions; mapping the feature map related to the approximate macular lesions to the lesion grading space through the classification layer; to output the prediction result of the degree of pathological myopia; the lesion grading space includes a grade space corresponding to leopard-like fundus, a grade space corresponding to diffuse choroidal atrophy, a grade space corresponding to patchy choroidal atrophy and a grade space corresponding to macular atrophy.

6. A readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for grading proximal maculopathy described in any one of claims 1 to 4 is implemented.

7. A grading device for proximal maculopathy, characterized in that: include: Processor and memory; The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the device for grading proximal maculopathy performs the method for grading proximal maculopathy as described in any one of claims 1 to 4.

Images